基于Logistic回归的区域地质灾害综合气象预警模型

Regional Integrated Meteorological Forecasting and Warning Model for Geological Hazards Based on Logistic Regression

应用Logistic回归拟合地质灾害发生概率,建立了区域地质灾害综合气象预警模型,实现了降雨引发地质灾害发生概率的动态预报。降雨观测和数值降雨预报作为模型的动态输入,利用信息量模型集成地学因子得到的总信息量作为模型的静态预报因子。预警模型在栅格点上预报降雨引发地质灾害的发生概率,并以等级形式发布预警。用2004年实际观测对预报进行检验,结果表明当年80%的灾害都达到了预警发布的标准。2006年汛期的业务运行对台风碧利斯引发的地质灾害做出了较准确的预报预警。说明模型能满足预警业务的要求,是提高地质灾害气象预警水平的有效途径。

Regional Integrated Meteorological Forecasting and Warning Model for Geological Haz- ards is constructed based on Logistic regression of the probabilities of geological hazard occur- rences. Dynamic forecasting of the probability of rainfall-induced geological hazard occurrence is thus realized. Rainfall observations and numerical rainfall forecasts are used as dynamical inputs of the model. Total information produced by information model that integrates factors of geosciences is used as the static predictor of the model. The model forecasts probabilities of geological hazard occurrences on a grid, and releases warning messages tion of the model with observational data for the year in the form of a 5-level hierarchy. Valida- 2004 shows that 80 % of the geolog-ical hazards of the year have been identified as Warning Level 3 or above. The operation of themodel during the rainy season of the year 2006 cy for geological hazards triggered by Typhoon requirements of warning operations, and is an ings for geological hazards. has made forecasts and warnings with high accura- Bills. All these show that the model can satisfy the effective way to improve the meteorological warn-